Thirty community-dwelling, moderately depressed elderly were randomly assigned to 1 of 3 interventions: experimenter-accompanied exercise in the form of walking, a social contact control condition, and a wait-list control. Exercise and social contact both resulted in significant reductions in both the total and the psychological subscale of the Beck Depression Inventory (BDI). The exercise condition, however, unlike the control conditions, resulted in decreased somatic symptoms of the BDI. These results indicate that, at least in the short term, exercise has a broader effect compared with control conditions in reducing depressive symptoms in the moderately depressed elderly.